Good points all around...

One of the things that bothered me the most when I was looking at some of these sources was the opacity of certain sites' methods (ahem, RCP). I couldn't figure out how in the hell RCP determined their categories, so I just eye-balled the chart below their map and guessed that, since there was no overlap between categories (minus the "Arizona asterisk"), there were hard-and-fast lines.

Being dense, however, I failed to notice that only around thirty states are actually listed in the chart. Oops.

The categorization of states in to Tossup / Lean / Strong will always be to a certain extent arbitrary--the only the only hope would be that, by including multiple sources, this arbitrariness averages out among them.

One reason why I like the inclusion of 538's numbers is that it avoids the potential problem inherent in categorizing states. As Leah and Scott have mentioned, 538 tends to over-use "leaners" relative to "tossups." While this may be true in the chart, it is nullified by the fact that the Forecast itself uses the 538 projection instead of the categories.

The same could be done with the InTrade numbers (I know, I know, I'll shut up about InTrade soon, I promise), with the Democratic (or Republican) price taken as the percent chance of that candidate winning that state, and using either the total expected value or the median result of simulation runs as the InTrade projection number.

True. I guess my point--poorly stated--was less that the current sources should be thinned than that they should be expanded to include a wider a diversity of data--namely, the InTrade market projections.

I feel like the addition of a source that is derived from neither polls nor pundits would add further breadth to the Forecast, making it a more interesting measure. The suggestion to cut certain sources was meant more as a "if you don't feel like checking ten sites, here's my feelings on the weakest links"-type idea.

So, first off let me just say that I love this site -- During the primary season, it became my Bible of sorts for getting the best / latest info on superdelegate positions and the Clinton-Obama horse race in general. Well done, site admins. You rock.

So now that I'm done sucking up, a couple things with regard to DCW's Presidential Forecast methodology/algorithm. As far as I can tell, the basic goal of the Forecast is to provide a relatively simple, up-to-date distillation of the info from many different sources, averaging them all into one nice number, complete with pretty map. Right now, the sources DCW is pulling from can be split into three basic types:

1. Simple poll averages

OpenLeft, EV.com, RCP, and FHQ are all nothing beyond (relatively) simple averages of polls. The time windows, weighting methods, and cut-off points (of what is considered "Strong" vs. "Lean", etc.) differ between the sites, but they're all still working from roughly the same playbook.

2. Polls plus magic

Elect. Proj. and 538 are averages as well, but include additional correcting factors in their methodologies that set them apart from the four sites listed above. Really, 538 deserves to be in a class of it's own, but I digress.

3. Polls, magic, hamsters, and the weather forecast, combined by secret formula (PMHWF-CbSF)

Rasmussen, CNN, and NBC have secret magical formulas that they use to derive their state rankings. The amount of objective vs. subjective information is really anybody's guess.
____
Anyway, my point is that using 4 of the 9 spaces for simple poll averages is a waste of space that could be better used for adding other, different data, making the Forecast a wider--and therefore more interesting/informative--measure. Since each site is weighted in the Forecast algorithm equally, including 4 simple poll averages has the additional problem of weighing this data too heavily with regards to the other sources.

Suggestion: Perhaps dropping two of the four, and adding InTrade market data?

And what is with the outrageous lameness of the never-updating CNN and NBC? As a child of the InterTubes, I demand instant information updated hourly! This old-media "meh, once a month is enough" mindset just won't fly. But it does seem like a good idea to include more than one PMHWF-CbSF in the Forecast, so if there aren't any other good PMHWF-CbSFs out there updating more frequently, I guess we're stuck with the infinite lame-itude of CNN/NBC.

But yeah, I don't mean to step on anyone's toes or anything--just suggestions. Let me know what you guys think.

Hey Leah, just thought I'd point you in the direction of some interesting commentary on the Ohio polls over at FiveThirtyEight (since I have nothing original to add myself):

http://www.fivethirtyeight.com/2008/07/todays-pol...
http://www.fivethirtyeight.com/2008/07/dueling-oh...

Worth a read, IMHO.